Side-channel analysis in frequency domain with clustering

Abstract

The alignment of power traces is one of the key factors affecting the success of side-channel analysis. Frequency-domain side-channel analysis can effectively solve the alignment of power traces in the time domain. However, frequency-domain side-channel analysis usually requires more power traces. In order to improve this problem, this paper proposes a frequency-domain side-channel analysis method based on clustering. After the effective signal frequencies are separated through the clustering algorithm in machine learning, side-channel analysis is carried out to reduce the number of power traces. Experiments show that this method can reduce the number of power traces required for frequency-domain side-channel analysis in different degrees.

Keywords

• side-channel analysis
• frequency domain
• machine learning
• clustering
• power traces